Tensor products on quiver representations. (Q2457289)

Around 1997, A. T. Vlassov and the reviewer initiated the study of the Clebsch-Gordan problem for the pointwise tensor product of quiver representations. Their original motivation was to understand the monoidal structures on representations of algebras arising from bialgebra structures on the algebras in question. This approach was illustrated using a toy example of the Jordan quiver (i.e., the quiver with one vertex and one loop). In that case the original problem becomes the problem of decomposing either the Kronecker (i.e., group-like) product of Jordan blocks or the primitive analog of the Kronecker product. This approach, inspired by quantum groups, cannot be used on pointwise tensor products without further modifications: if the quiver has more than one vertex, then the pointwise product is never induced from a bialgebra structure -- this can easily be seen by computing the dimension vector of the representation space. Said differently, the group-like comultiplication is not an algebra homomorphism if the quiver has two or more vertices. In the paper under review the author provides a ``comparative study'' of the pointwise tensor product vs. the group-like product on a path algebra \(A\). That product does not, in general, result in a bialgebra structure on \(A\) and, consequently, one is lead to consider what the author calls \(A\)-premodules. He shows that any \(A\)-premodule is a direct sum of two unique subpremodules, one of which is an \(A\)-module. It is then shown that when the premodule is the group-like product, the unique \(A\)-module summand of it is exactly the pointwise tensor product. This is done in Section 2. In Section 3, a different approach is presented. To this end, the author describes the tensor product (over the ground field) of two path algebras as the path algebra of a quiver with relations. Furthermore, for bialgebra structures he does not require coassociativity or the existence of a counit, and he specializes further to the comultiplication such that \(\Delta(a)=a\otimes a\) for any arrow \(a\) in the quiver. A further compatibility assumption is then made: a linear map \(\Delta\) from the algebra to its tensor square is said to be partitioning if \(\Delta\) applied to any vertex \(e_k\) is the sum of tensor products of vertices, and the corresponding pairs of indices give rise to a partition, indexed by \(k\), of the Cartesian square of the set of the vertices. The author then shows that if the comultiplication is a partitioning algebra homomorphism under which each arrow is group-like, then the \(\Delta\)-induced tensor product of two representations decomposes as the direct sum of the pointwise tensor product of the two representations plus a direct sum of simple modules with explicitly described multiplicities. In particular, the Clebsch-Gordan problems for the two tensor products are equivalent.